Shape-adaptive robust regression¶
Conceptual illustration; not an estimated residual distribution.
Python estimators for regression under skewness and heavy tails, developed for the Arashi–Yüzbaşı study From Residual-Moment Penalties to Shape-Adaptive Robust Estimation under Skewness and Heavy Tails.
Choose an estimator¶
| Need | Public interface | What to inspect |
|---|---|---|
| Shape-adaptive elastic net | ShapeAdaptiveEN |
Pilot, bounded-score iteration and CV/refit diagnostics |
| Shape-penalized elastic net | ShapePenalizedEN |
Shape weights, robust option and convergence output |
| Quantile comparison | QuantileLasso |
Quantile levels and optimization success |
| Training-data tuning helpers | fit_en_cv, fit_sken_cv, fit_quantile_lasso_cv |
Candidate grids, folds and seeds |
| Numerical summary | diagnostic_summary |
Final KKT residuals and iteration-cap counts |
Clear boundaries¶
Version 0.1.0 provides the installable estimator API and a small installation
test. This is a GitHub distribution, not a PyPI release. It requires
Python 3.13 and the exact numerical dependency versions in pyproject.toml.
The paper's application data, simulation scripts, generated results and manuscript are not included in this repository. The original providers and research-verification route are described under data and reproducibility. No benchmark or general performance guarantee is implied by this showcase.
From software to research¶
The citation and credit page records authorship and code provenance. Publication details will be added when verified; no journal DOI is asserted here. Return to the author's software catalogue to explore related R and Python packages.